An Improved Hilbert-Huang Transform And Its Application In Feature Extraction

Signal analysis is the main means of feature extraction method of the fault characteristic signal. So, it is important to research advanced non-stationary signal process method to extract fault features accurately. Hilbert-Huang Transform(HHT) proposed by scientist Norden E Huang is a new time-frequency signal analysis method. The algorithm is very suited for processing nonlinear and non-stationary signals, and good results have been achieved when it is applied to mechanical fault diagnosis, engineering,and biomedical science and so on. The algorithm has great theory research value and widely used in engineering filed. Since HHT is still in the preliminary tages of development. Thus there are a number of issues need to be studied and improved, like curve fitting, border processing. Therefore, this paper analysis and discuss the problem of the HHT, and proposed measures for improvement. Its contributions list as follows:①On the basis of a great deal of researches in the past about the theory and applications of Hilbert-Huang Transform, effors are made tring to improve and develop its theory. The fitting of the envelop line and mean curve line is a key problem of the HHT, so the study on this problem is very important. This paper summatizes the shortcomings of the cubic spline interpolation, at the same time, the shortage of ordinary end extending methods is pointed out and the reason of end effect is analyzed. Summarized the exixting boundary process methods.②The shortage of cubic spline interpolation is pointed out, and proposed an additional interpolation, B-spline method .Apply this method to improve the decomposition efficiency and reduce the computation. the simulation and practical signals analysis show that the proposed method is better than the EMD based on cubic spline interpolation.③The shortage of ordinary end extending methods is pointed out and the reason of end effect is analyzed. A novel method for restraining the end effects of BS-EMD based on the support vector machines is proposed. The simulation and practical signals analysis show that the proposed method is better than other methods.④Based on the above researches, a HHT analysis system for non-stable signal is developed using the object-oriented programming language VC++. The system includes of time-domainsis. ⑤The improved HHT method is used to analyz fault vibration signals of rotating machinery. The Hilbert spectrums of rotor running across oil film instability fault are given respectively. The oil film instability fault analysis show that the proposed method can extract fault features.